Cognitive and Affective Aspects of Thigmotaxis Strategy in Humans

Janos KallaiUniversity of Pecs

Tamas MakanyUniversity of Southampton

Arpad Csatho, Kazmer Karadi, David Horvath,Beatrix Kovacs-Labadi, and Robert Jarai

University of Pecs

Lynn NadelUniversity of Arizona

Jake W. JacobsUniversity of Arizona and University of Arizona South

The present article describes the cognitive and emotional aspects of human thigmotaxis (a wall-followingspatial strategy) during exploration of virtual and physical spaces. The authors assessed 106 participantswith spatial and nonspatial performance-based learning–memory tasks and with fear and anxietyquestionnaires. The results demonstrate that thigmotaxis plays a distinct role at different phases of spatiallearning. The 1st phase shows a positive correlation between thigmotaxis and general phobic avoidance,whereas there is no association between thigmotaxis and general phobic avoidance during later phases oflearning. Furthermore, participants who underperformed in working memory tests and in a spatialconstruction task exhibited greater thigmotaxis and a higher potential for fear response. Findings areinterpreted in the framework of interactions among emotion-, action-, and knowledge-controlled spatiallearning theories.

Keywords: thigmotaxis, spatial exploration, cognitive map, fear, anxiety

Research on the analysis of spatial strategies draws a distinctionbetween action- and knowledge-based spatial orientation (Hartley,Maguire, Spiers, & Burgess, 2003). Thigmotaxis, the way in whichan organism organizes behavior relative to tactile stimuli, is aphylogenetically old and energetically inexpensive exploratorystrategy that embodies both action- and knowledge-based explo-rations (Creed & Miller, 1990; Fraenkel & Gunn, 1961). As such,thigmotaxis provides an opportunity to examine relations amongemotion, action, knowledge, and the temporal dynamics of spatiallearning.

Stimulus–Response Route Following

Action-based route-following strategies involve a behavioralsequence based on an egocentric frame of reference that—through

reinforcement, punishment, or extinction—associates cues withactions in a place and time. In guidance-based navigation (O’Keefe& Nadel, 1978), spatial orientation is composed of actions ac-quired during an experiential learning history, beginning at adefined start position (usually identified by the presence of a locallandmark), which triggers a specific response sequence leadingfrom the start position to a goal (Reid & Staddon, 1998).

Within this system, generalized goal-relevant actions includeappropriate responses to a localized stimulus cue and the sequen-tial use of navigational landmarks associated with the action.Although learning obviously occurs, the organism does not inte-grate these local landmarks into a system that includes spatialrelations; instead, it encodes only sequential associations betweencues and actions.

Models examining route following do not explain how an or-ganism acquires spatial knowledge; instead, they describe specificaction sequences. These models explain why numerous spatialerrors occur during route following (e.g., Reid & Reid, 2005;Wang & Spelke, 2004). Generally, these models claim that sucherrors are due to limited memory capacity for successive actionpatterns occurring under novel circumstances at different timesand in different contexts.

Egocentric models of route following (e.g., Gallistel, 1990;Hartley et al., 2003; Wang & Spelke, 2004) rely on distinctmechanisms to account for spatial-orientation errors that occurmainly in large scale spaces. First, the models invoke a dynamicupdating of previously visited landmarks and completed actions.Second, the models propose that, during updating, the occurrenceof integration error between the selected path and action patterns isnecessary (Gallistel, 1990; Whishaw, Hines, & Wallace, 2001).

Janos Kallai, Arpad Csatho, and Kazmer Karadi, Institute of BehavioralSciences, University of Pecs, Hungary; Tamas Makany, School of Psy-chology, University of Southampton, United Kingdom; David Horvath,Beatrix Kovacs-Labadi, and Robert Jarai, Institute of Psychology, Univer-sity of Pecs; Lynn Nadel, Department of Psychology, University of Ari-zona; Jake W. Jacobs, Department of Psychology, University of Arizona,and Department of Psychology, University of Arizona South.

The study was supported by Orszagos Tudomanyos Kutatasi Alap(Hungarian Scientific Research Foundation) Grant T-42650. We thankJames Comstock and William Cook (informally known as William/James)for writing the software for the computer-generated arena and graciouslymaintaining it for our laboratory and the laboratory of others.

Correspondence concerning this article should be addressed to JanosKallai, Institute of Behavioral Sciences, University of Pecs, 12 SzigetiStreet, Pecs, Hungary H-7624. E-mail: janos.kallai@aok.pte.hu

Behavioral Neuroscience Copyright 2007 by the American Psychological Association2007, Vol. 121, No. 1, 21–30 0735-7044/07/$12.00 DOI: 10.1037/0735-7044.121.1.21

21

In these models, the temporal organization of successive eventsorganizes behavior. Hence, these models adequately describe theacquisition and execution of a path (route) in restricted and famil-iar space, but they do not describe how an organism orients in anew space, finds short cuts, or finds alternative routes when newobstacles arise along the path route (Reid & Reid, 2005).

Cognitive Exploration

In contrast, knowledge-based models of spatial behavior postu-late the existence of a mental construction that serves to mediateinteractions among behavior, goals, and an integrated system ofnavigational objects (cues and/or landmarks). Theoretically, thismental construction aids in the organization of a coherent view ofspace and relates that view to sets of latent behavioral strategies,each of which might be used to search the space or to locate and/orrelocate places (e.g., goals) in that space. This mental constructionconsists of either a viewer-dependent apperception of the geomet-ric properties of the environment (e.g., Cheng, 1986) or of aviewer-independent abstract knowledge system built from variouscognitive modules that construct and upload coherent spatial in-formation (O’Keefe & Nadel, 1978). Hence, these approachesdescribe an abstract mental system that builds up contextual in-formation over time but limits the importance of temporal infor-mation itself. This context-specific spatial knowledge may serve asa navigational guide by providing detailed information about thelocation of relevant goals within a coherent system of relatedsalient navigational landmarks. Nevertheless, it is currently a mat-ter of intense theoretical debate regarding how context-specificspatial knowledge translates into coherent spatial navigation(Wang & Spelke, 2004). Thigmotaxis, an automatic, biologicallyprimitive strategy for exploring space, may provide a lens throughwhich the nature of this dynamic processing in humans’ spatiallearning may be revealed.

Thigmotaxis

Although thigmotaxis is a well-characterized behavioral tacticcommonly observed in nonhuman animals, its role in humannavigation is yet to be explained. Etymologically, the term thig-motaxis is based on the Greek word thigma, meaning contact withan object. The expression taxis refers to the reaction of an organ-ism to external stimuli by movement in a specific direction. Whenan animal initially explores an enclosed place, it tends to stay inclose contact with the perimeter of that space (Barnett, 1968). Onemay quantify the tendency to avoid the inner zone of an open fieldby measuring either the time or the path length that an organismspends in close contact with the wall. This behavior is sometimesknown as wall-following, wall-touching, or centrophobic behavior(Besson & Martin, 2004; Creed & Miller, 1990; Fraenkel & Gunn,1961). Thigmotaxis is especially prominent on the initial encounterof a novel space, is a dominant factor early in the exploration of anenclosed space, and is not associated with the rate of acquisition orquality of retrieval of spatial knowledge (Choleris, Thomas, Kava-liers, & Prato, 2001). Instead, thigmotaxis appears to help anorganism define the boundary of an enclosed space and serves asa “home base” from which the construction of a spatial map mayoccur. Obviously, thigmotaxis permits the organism to determinethe boundaries of an enclosed space; its continued use, however,

prevents other spatial strategies that permit the formation of acognitive map of an environment. Hence, the continued use ofthigmotaxis prohibits an organism from locating or relocating anescape platform during the middle and later phases of spatiallearning in a Morris-type maze task (Graziano, Petrosini, & Bar-toletti, 2003; Kallai, Makany, Karadi, & Jacobs, 2005).

Damage to the lateral caudate putamen influences many of theexploratory strategies exhibited by participants in spatial learningexperiments (McDonald & White, 1994), yet a primary effect is adramatic increase in thigmotaxis. Specific lesions to these dorso-medial striatal areas result in an impairment of acquisition andexpression of learned responses to spatial cues (Devan, McDonald,& White, 1999).

Thigmotaxis is frequently found in highly anxious animals(Jeanson et al., 2003; Ohl, Sillaber, Binder, Keck, & Holsboer,2001; Pellow & File, 1986). Laboratory investigations studying theeffects of anxiolytic drugs reveal that thigmotaxis is one manifes-tation of a biologically prepared fear reaction and that it plays animportant role in the formation of anxiety-induced avoidancebehavior and cognition. Some argue it may serve as a model for theanalysis of human anxiety disorders (Treit & Fundytus, 1988).

Thigmotaxis as defined above is a primordial, geneticallygrounded type of behavior and an ecologically important strategyused by humans and other animals for spatial exploration (Creed &Miller, 1990; Jeanson et al., 2003; Kallai et al., 2005; Simon,Dupuius, & Constentin, 1994) mainly in open-field and arena–maze experiments. Although studies with nonhuman animals de-scribe thigmotaxis behaviorally, its relationship to human cogni-tive or affectively driven performance remains to be determined.

Current research in animals and humans stresses that novelspatial situations trigger thigmotaxis. It is typically manifestedwhen searching boundaries of an enclosed space to find a startingpoint (home base) for further spatial exploration. Its presence alsonecessarily delays or inhibits the exploration of the central zone ofa novel space until the organism identifies salient or relevant cues(Besson & Martin, 2004; Kallai et al., 2005). Once identified, thesesalient or relevant cues serve as elements of a spatial or cognitivemap. In the early phase of map construction, the salient or relevantcues are not integrated into a cognitive map; instead, they are onlycues with emotional significance for approach or avoidance.

Although measures of thigmotaxis taken during the first phaseof exploration do not predict the eventual success of spatial learn-ing (Lipp & Wolfer, 1998), thigmotaxis may serve as an essentialelement in the control of the dynamics of spatial learning. Themajor transient point of a series of learning trials that can predictthe success of spatial acquisition occurs during the middle trialswhen a home base has been established, egocentric space calibra-tion has occurred, and the acquisition of the structure of Euclideanspace has begun (Kallai et al., 2005).

Fear and Anxiety

The strategies that an organism uses to explore a novel envi-ronment are guided by behavior economy and factors related toindividual differences in spatial neophobia, the organism’s per-ceived ability to escape from novel environments, and the organ-ism’s tolerance for ambiguous situations. Spatial anxiety (Lawton,1994), sometimes known as spatial neophobia, and the fear oflosing one’s way (Kozlowski & Bryant, 1997) differ fundamen-

22 KALLAI ET AL.

tally from anxiety in humans. Fear and neophobia are related tolow fear-response threshold in fear-provoking situations, originat-ing from inborn escape-reaction sensitivity (Kagan et al., 1990;Klein, 1981). The appearance of these types of fear (some mightcall it caution) is marked by elevated safety-seeking strategies thatinfluence (a) actual behavioral sequences (Jeanson et al., 2003), (b)tactics used to explore unfamiliar places (Bryant, 1997), (c) themaintenance of attentive focus on the geometrical features of theenvironment (Kozlowski & Bryant, 1997), (d) the ability to mem-orize spatial locations (Evans, Skorpanich, Ganling, Bryant, &Bresolini, 1984), (e) the speed and fluency of the navigationmovements and the mobilization of configuration knowledge(Schmitz, 1997), and (f) the integration of ego-allocentric-basedspatial information (Jacobs & Nadel, 1999; Kallai, Kosztolanyi,Osvath, & Jacobs, 1999).

In light of several methodological complexities, we assessed acarefully selected sample of participants with different levels offear and anxiety. Our purpose was to examine cognitive andemotional factors that may underpin thigmotaxis in virtual andphysical arena mazes and in different spatial and nonspatial learn-ing and memory tasks. First, because thigmotaxis plays differentroles in the various phases of spatial learning in humans, our aimwas to identify emotional and cognitive factors underlying the useof thigmotaxis during different phases of spatial learning. Second,we attempted to determine the statistical relations among measuresof the identified emotional and cognitive factors and of the man-ifestation of thigmotaxis. Finally, we characterized the role thig-motaxis plays in human orientation and way finding in a novelenvironment.

Method

Participants

A total of 106 participants took part in the present study.Thirty-nine male participants, ranging in age from 19 to 26 years,with a mean age of 21.41 years (SD � 1.8), and 67 femaleparticipants, ranging in age from 19 to 26 years, with a mean ageof 21.40 years (SD � 1.5), were recruited through advertisementsand received compensation for participating. The experimenterdescribed the experimental procedures as a part of the informedconsent procedure. No participant had previous psychiatric ill-nesses or physical disabilities that could interfere with the com-pletion of the computer-based or physical spatial tasks. Anticipat-ing a possible distortion coming from individual differences ofcomputer gaming experience (Waller, 2000), participants withself-reported computer game playing time exceeding 30 min/weekwere ineligible for the study. The study procedures met or ex-ceeded standards established by the Helsinki Declaration of 1975,as revised in 1983.

Apparatus

Computer-generated arena (CGA). A desktop-based com-puter program was used to record and measure search strategiesthat participants used to locate and relocate a specific place in acomputer-generated space (for details see Jacobs, 1997; Jacobs,Laurance, & Thomas, 1997).

The participants viewed a monitor, attached to a standard PC,which displayed a circular arena located within a square room.

Each wall in the room had a distinctive texture and contained localicons consisting of windows or arches, which provided means bywhich participants could orient themselves within the virtualspace. The participants viewed the arena from a first-person per-spective as though standing or moving on the floor of the arena.

Real arena maze (RAM). A large-scale physical environmentwas used to record and measure exploration strategies. This RAMwas built to model the Morris-type maze task (Morris, 1984) inhumans. The RAM task was first developed and described byKallai et al. (2005; see Figure 1 for a picture and see http://www.aok.pte.hu/spacelab/ for further details).

The RAM apparatus consisted of a large, circular timber wallarena (6.5 m in diameter, 2.0 m in height, 33.0 m2 in floor space,with an inner temperature of 20 � 5 °C). Inside the arena, eightnavigation objects were placed on 1-m high shafts. Each object haddifferent fixed geometric cues that were unique in shape but equalin size and surface.

Three starting positions (A, B, C) were located equidistantaround the edge of the west and south wall of the arena. A roundtarget platform disc (50 cm in diameter) was placed in a fixedlocation within the northeast quadrant of the maze floor and at anequal distance from the three starting positions. The target wasequipped with pressure-sensitive detectors. When the participantstepped on the target platform, a 60-dB tone was emitted. Theparticipants received no other feedback on their performance. Thelocation of the navigation cues and the target remained fixedacross trials.

The RAM was equipped with doors on the west and south walls.The doors served as an entrance or an exit on a pseudorandomschedule. Before entering the RAM, participants were fitted withopaque swimming glasses, so that they could not obtain any visual

Figure 1. A bird’s-eye view perspective image taken from the real arenamaze depicting the entrance and exit doors, the eight navigation objects,and the circular platform (target) on the floor. The three starting positionsare indicated with white letters (A, B, C).

23THIGMOTAXIS STRATEGY IN HUMANS

information on the layout that might have helped them to locatelandmarks or the target. Spatial performance was recorded andscored by the EthoVision 3.0 video tracking system (Noldus,Spink, & Tegelenbosch, 2001).

Procedure

CGA. The experimenter seated participants in front of a 17-in.(43.18-cm) super video graphics array monitor, a pair of stereospeakers, and a joystick controlled by a standard desktop PC andread a set of standardized instructions (see Appendix). The CGAsoftware displayed two types of rooms. The first was a practiceroom with no objects or platforms in it. The purpose of this roomwas to familiarize participants with the virtual environment andpractice virtual locomotion in that environment. After becomingfamiliar with a virtual environment, the participants were tele-ported to the test room. The test room consisted of a circular arenawithin a bigger square room. The room had a floor with anirregular texture and a gray ceiling. Distal objects were placed onthe walls. A blue rectangular target appeared on the floor (seeTable 1 for a description of the parameters used in the practice andtest rooms).

The target was visible in the first two probe trials (Probe 1 andProbe 2). In the test phase from Trial 1 to Trial 8, the target wasinvisible, though it remained in the same fixed location. Theparticipants’ task was to locate and relocate the target as quickly aspossible on each trial. The position of the textured walls and thedistal objects on them remained constant across all trials.

Recorded movement in CGA. The CGA software recordedmovement in the virtual environment as a set of binary variables(vector coordinates) sampled at a rate of 35 frames/s. Hence, the

software recorded virtual motion as a change in these coordinatesacross time. The CGA also calculated the latency to locate thetarget, the path length from the start position to contact with thetarget, and the path maps, which could be presented in a bitmapformat from the full set of vector coordinates generated withineach trial. These CGA calculations served as the basis of thestatistical analyses.

Identifying thigmotaxis. Thigmotaxis, in the present context, isthe portion of the path map that represents movement close to thearena wall (see Figure 2A). This is a summed value of every suchmotion pattern in the actual trial. In many cases, thigmotaxisreoccurred at the same section of the arena wall, so the finalsummed value may exceed 360° (the total circle of the arena wall).We measured thigmotaxis navigation strategies (represented in thefollowing equations as CGATHIGM) in degrees and then trans-formed the value into Arena Units (virtual meters; vm). We cal-culated vm using the following formula:

CGATHIGM ��

180* wall° * rCGA

where wall° is the degree of the measured overall distance betweenthe initial wall touch and the terminal wall touch, and rCGA is theradius of the CGA in Arena Units (50 Arena Units in this exper-iment).

To investigate temporal dynamics of thigmotaxis, we deter-mined four different measures by collapsing individual trials intoblocks. These blocks were as follows: first trial (Trial 1), earlyphase (the mean of Trials 2, 3, and 4), middle phase (the mean ofTrials 5 and 6), and late phase (the mean of Trials 7 and 8).Furthermore, an overall measure was also calculated as the meanof the scores from all the trials (overall CGA thigmotaxis).

RAM. An assistant guided the participants into the RAM via apseudorandomly chosen entrance door to either of the three start-ing positions (see Figure 1). The assistant instructed the partici-pants to locate the target using the tactile navigational cues, whichwould help them find and learn the shortest route towards theplatform. The assistant also instructed the participants to payparticular attention to the spatial relations among the landmarksafter locating the escape platform and to remember the spatialrelations among the navigation cues.

Each acquisition trial lasted a maximum of 300 s. If the partic-ipant failed to locate the target within 300 s, then a new trial wasstarted. If the participant managed to locate the target, then anextra 15 s were given to provide time to explore the surroundingenvironment. There were seven consecutive acquisition trials with2-min intertrial intervals. After completing the trial, an assistantguided the vision-occluded participant out of the maze.

RAM data collection. The variables included a measure of thecontact with the arena wall in meters from trial to trial(RAMTHIGM). Following the CGATHIGM calculation describedabove, RAMTHIGM was transformed into meters for each trial.The participants’ performance on each trial was recorded by avideo camera fixed above the center of the arena. Two independentraters assessed thigmotaxis manually. Following the participants’path along with contact of the wall, the raters used a protractor toscore thigmotaxis (see Figure 2B). The RAMTHIGM for each trialwas transformed into meters using the formula:

Table 1The Computer-Generated Arena (CGA) Structural and MotionParameters

Parameter Measurement

Room

Room dimensions 512 � 512 � 128Arena wall radius 50Arena wall height 3.5Target dimensions 10 � 10

Participant

Participant eye heighta 2.0Field of view 50°

Motionb

Move quantum 0.80Turn quantum 1

Note. Unless otherwise indicated, all measurements are in Arena Units.The reader may consider an Arena Unit as the equivalent of a virtual meter.a The PC monitor displays all views of the arena from a first-personperspective and as if the eyes of the participant are a certain user-defineddistance (an “eye height”) from the floor of the CGA rooms.b The move quantum describes the shift in the participant’s view of theCGA with each stroke of the up or down arrow. Similarly, the turnquantum describes the shift in the participant’s view of the CGA with eachstroke of the right or left.

24 KALLAI ET AL.

RAMTHIGM ��

180* wall° * rRAM

where rRAM � 3.25 m, which was the radius of the RAM.For analyzing the temporal dynamics of thigmotaxis in the

RAM, we used the same four phases as described in the case of theCGA. The phases were first trial (Trial 1), early phase (the meanof Trials 2, 3, and 4), middle phase (the mean of Trials 5 and 6),and late phase (Trial 7). Furthermore, an overall measure was alsocalculated as the mean of the scores from all the trials (overallRAM thigmotaxis).

Map construction posttest. After leaving the RAM, the partic-ipants were asked to construct a 3:1 scaled-down model of theRAM on a round table. The participants were provided with eightminiature navigation cues similar to those that were used in theRAM and with eight distractor objects with small shape differ-ences from the original cues. They were asked to select theappropriate cues and construct an accurate model of the RAM and

to mark the place of the target. The set of miniature geometricobjects provided no information on the spatial order of the RAM.

A relative distance of the target from its correct location (plat-form deviation) on the scaled-down posttest arena was scored withthe following method: The placement of the escape platform wasfitted in an x, y coordinate system, and the distance from the properlocation (measured in the same x, y coordinate system) was used asan error score.

Fear and anxiety. A self-reported test battery including theFear Survey Schedule (FSS; Arrindell, 1993) with Social Fear,Agoraphobia, Fear from Sexual and Aggressive Scenes, Fear ofIllness subscales and the overall fear-driven Avoiding BehaviorActivity (Overall Fear) score was used. The participants’ anxietylevels were assessed by the anxiety scale of the Spielberger’sState–Trait Anxiety Inventory (STAI; Sipos, 1978; Spielberger,Gorsuch, & Lushene, 1970). See Table 2 for a summary of thesescales and the main scores. Gender-related differences in thescores are indicated and discussed.

Figure 2. A: The circular computer-generated arena (CGA) is shown in a plane view with an imaginarydivision into four quadrants. The location of the platform is indicated on these maps as a square, and thenavigation path is indicated with a continuous line. B: The circular real arena maze is shown in a plane view withthe same imaginary divisions into the four quadrants as in the CGA. The division of inner and outer zones isshown with dotted lines. The navigation search path is indicated with a continuous line. T:1–T:8 � Trials 1–8.

25THIGMOTAXIS STRATEGY IN HUMANS

Measurement of cognitive functions. The subscales of theHungarian version of the Wechsler Adult Intelligence Scale(WAIS–H; Kun & Szegedi, 1971; based on Wechsler, 1997) wereused to assess the participants’ cognitive functions. The WAIS–Hstandard scores were calculated from raw scores according to thestandards of the Hungarian manual (Kun & Szegedi, 1971). Thereare 10 subtests in the WAIS–H: Information (general knowledgerecall), Comprehension (adaptation to everyday situations), DigitSpan (numerical working memory), Arithmetic (arithmetic skills),Similarities (abstract reasoning), Digit–Symbol (perceptuo–motorskills), Picture Arrangement (understanding social interactions),Picture Completion (closure and goodness of gestalt formation),Block Design (spatial construction skills), and Object Assembly(spatial integration skills).

Results

We performed two main analyses to examine the relationship ofaffective and cognitive components with the occurrence of thig-motaxis strategy in each trial of the CGA and RAM tasks. Apreparatory examination on the data with independent-sample ttests indicated a significant gender effect in the self-reportedquestionnaires, as women scored higher than men on all subscalesexcept for Fear from Illness (see Table 2). Consequently, allfurther analyses considered gender differences.

Analysis of the Affective Components (Fear and Anxiety)

An analysis using partial correlation coefficients was carried outto examine the relationship between affective components (fearand anxiety) and thigmotaxis in the two test environments (CGAand RAM). Gender was used as a controlling variable in all cases.See Table 3 for a summary of the findings.

CGA. The measure of Overall Fear correlated with thigmo-taxis in the early phase of CGA, r(104) � .22, p � .05. The otherphases were not associated with fear. In addition, Overall Anxietydid not correlate significantly with any of the CGA thigmotaxisscores.

RAM. Overall Fear correlated with thigmotaxis in the first trialof RAM, r(103) � .26, p � .05. However, other phases did notshow significant associations with the measure of fear. Similarly,as in the case of CGA, there was no correlation between OverallAnxiety and RAM thigmotaxis.

Taken together, it appears that the psychometrically assessedlevel of fear correlated with high rates of thigmotaxis in the initialphases of place learning but that the anxiety scores do not correlatewith measured thigmotaxis at any point in learning.

Analysis of the Cognitive Components

A second analysis focused on how cognitive factors mightinfluence the occurrence of thigmotaxis strategy in different

Table 2Descriptive Statistics and Independent Samples t-test Statistics for the Gender Split of the FearSurvey Schedule (Arrindell, 1993) and the Spielberger’s State–Trait Anxiety Inventory’s AnxietyScale (Sipos, 1978; Spielberger, Gorsuch, & Lushene, 1970)

Questionnaire and subscale

Male Female t test

M SD M SD t df

Fear Survey ScheduleSocial Fear 14.82 9.70 22.31 9.31 3.94** 104Agoraphobic Fear 6.54 4.78 15.85 6.42 7.87** 104Fear from Sexuality 4.26 3.45 9.11 4.69 5.63** 104Fear from Illness 9.79 5.76 11.79 9.02 1.24 104Overall Fear 4.15 4.57 9.78 5.88 5.13** 104

Spielberger’s State–Trait AnxietyInventory

42.05 7.59 46.84 0.84 3.07* 104

* p � .05. ** p � .001.

Table 3Summary of the Partial Correlations Between the Overall Affective Components (Fear and Anxiety) and the Thigmotaxis Measured inthe Two Environments (CGA and RAM)

Component

CGA RAM

Firsttrial

Earlyphase

Middlephase

Latephase Overall

Firsttrial

Earlyphase

Middlephase

Latephase Overall

Fear .00 .22* .13 .11 .17 .26* .11 �.15 �.08 .13Anxiety .11 .15 .14 .08 .16 .12 .00 �.03 �.05 .04

Note. Results are controlled for gender. CGA � computer-generated arena; RAM � real arena maze.* p � .05.

26 KALLAI ET AL.

phases of spatial learning. Categorical grouping variables werecreated on the basis of the 33rd and 66th percentiles of the 10subtests from the WAIS–H and of the platform deviation measureof the map construction posttest. These grouping variables wereused as independent variables. Furthermore, gender was includedas a covariate factor to control for sex differences. Two multivar-iate general linear models were used to examine the effects of theindependent variables on the occurrence of thigmotaxis strategy asa dependent variable within the two environments (CGA andRAM).

CGA. Participants with extensive thigmotaxis usage (M �107.59 vm, SE � 18.62) in the early phases had scored signifi-cantly lower (�14 points) on the Information subtest than did theparticipants who showed less thigmotaxis (M � 44.04 vm, SE �24.97) but had higher Information scores (�16 points). The effectwas significant, F(2, 105) � 3.49, p � .05. Similarly, in the BlockDesign subtest, the higher scoring (�15 points) participants usedless thigmotaxis (M � 38.14 vm, SE � 19.07) in the early phasethan did the lowest scoring group of participants (�14 points) withintense thigmotaxis (M � 110.92 vm, SE � 22.26). This was alsoa significant effect, F(2, 105) � 3.50, p � .05. Early phase andoverall thigmotaxis were also significantly different in the threecategorical groups of platform deviation, F(2, 105) � 6.40, and,F(2, 105) � 4.81, respectively (both ps � .05). In the early phase,participants with good platform-position-recall ability (M � .98cm) had less thigmotaxis (M � 55.54 vm, SE � 17.31), whereasparticipants with higher platform deviations (M � 13.96 cm) hadmore thigmotaxis (M � 116.96 vm, SE � 18.52). Similarly,participants with high overall thigmotaxis (M � 725.25 vm, SE �108.31) showed more platform displacement in the posttest thandid participants with low overall thigmotaxis (M � 377.61 vm,SE � 101.24) who were more precise with remembering thelocation of the platform.

RAM. Participants who used thigmotaxis extensively (M �13.84 m, SE � 2.86) in the early phase of spatial learning scoredhigher (�16 points) on the Information subtest than did those whoused less thigmotaxis (M � 6.58 m, SE � 2.07; �14 points on theInformation subtest). This effect was significant, F(2, 105) � 3.76,p � .05. Reversely, participants with low scores on the BlockDesign subtest (�14 points) had extensive thigmotaxis (M �2.96 m, SE � 1.44) in the middle phase, whereas those with highscores (�15 points) had low thigmotaxis (M � .593 m, SE �1.24). This effect was also significant, F(2, 105) � 3.27, p � .05.Although all the cognitive ability variables were included in eachmodel, for the sake of clarity we only present the significantvariables in Table 4.

The final data analysis examined the relationships between theaffective and cognitive measures. A regression analysis demon-strated that overall fear could be moderately predicted by theworking memory dependent platform deviation variable, R2 �.210, F(11, 104) � 2.25, beta � .277, t(105) � 2.70, p � .05.However, the same relationship was not found in the case ofanxiety.

Discussion

Studies with mammals, insects, and worms indicate that thig-motaxis is a strategy used by organisms seeking safety (Besson &Martin, 2004; Choleris et al., 2001; Creed & Miller, 1990; Treit &

Fundytus, 1988). This safety-seeking strategy appears in the earlyphases of exploration in new and potentially dangerous situations.The present study investigated differences in thigmotaxis activityin a visually dominated CGA and a perceptuo–motor-guided RAMduring the first, early, middle, and late phases of spatial learning inhumans. The data analyses suggest several principles.

First, humans who use egocentric thigmotaxis strategy duringthe early trials of virtual or real mazes also exhibit high levels ofpsychometric fear and avoidance bias for fear-mobilizing situa-tions. This suggests that human thigmotaxis, the tendency torefrain from exploring the inner zone of novel places, is similar tothat exhibited by other animals (Besson & Martin, 2004; Treit &Fundytus, 1988)—its presence indicates a general bias to cautious,safety-seeking phobic behavior.

Second, limits in working memory and gross mapping errors areparts of an enhanced use of thigmotaxis. We suggest that a deficitin general spatial integration plays a significant role in maintainingthigmotaxis activity, principally during the early to middle phasesof spatial learning. This means that humans with spatial integrationdifficulties are cautious navigators; they use thigmotaxis moreextensively in the early and middle phases of place learning. Theyalso tend to avoid the central zone of an enclosed space and preferto stay close to the border for a longer period of time.

Third, the results of the analysis of the RAM data revealed apattern showing that, similar to the CGA task, deficits in spatialintegration and an extensive use of thigmotaxis are related duringthe early and middle phases of spatial learning. These resultssupport the proposal that the capacity to form gestalts plays a rolein switching from egocentric navigation to allocentric-based mapconstruction. Consequently, deficits in spatial gestalt constructionincrease the likelihood of egocentric-based thigmotaxis behavior.

The question that arises is as follows: How do phase andmechanism tip the balance between avoiding and embracing ex-ploration? We suggest that the individual differences in this point

Table 4Summary of the F Values for the Information, Block Design,and Platform Deviation Scores on the Thigmotaxis Measures

PhaseInformation

Subtest

BlockDesignSubtest

Platformdeviation

CGA, F(2, 105)

First phase 0.28 1.23 1.66Early phase 3.49* 3.50* 6.40*

Middle phase 1.04 0.53 0.93Late phase 0.57 0.93 1.88Overall 1.00 2.18 4.81*

RAM, F(2, 105)

First phase 1.65 1.41 1.77Early phase 3.76* 0.39 0.78Middle phase 0.41 3.27* 0.75Late phase 0.76 0.33 0.47Overall 2.79 1.10 1.19

Note. Gender was used as a covariate. All values without asterisks arenonsignificant. CGA � computer-generated arena; RAM � real arenamaze.* p � .05.

27THIGMOTAXIS STRATEGY IN HUMANS

vary because of the ability of the participants to construct andcomplete the basic structure of the current surroundings. If a goodgestalt of a given space is not constructed, then thigmotaxis re-mains active and the basic egocentric spatial definition prevails.The main condition for the switch between egocentric and allo-centric representations would be whether the individual forms anintelligible gestalt (cognitive map) of the surroundings or not.

Fourth, participants with a high level of thigmotaxis activitymanifest a low rate of spatial-map-formation ability and largerelocation errors during the early and middle phases of learning torelocate the hidden platform. Large relocation errors on the min-iature map model of the RAM indicated that participants with ahigh thigmotaxis score estimated the distance from the arena wallto the hidden platform to be greater than it is in reality. This biasin distance estimation appears with extensive use of CGA thigmo-taxis during the early phase of spatial learning.

The regression model demonstrated that placement error corre-lated positively with measures of phobic avoidance but not withhigh levels of anxiety. Our results support previous propositions(e.g., Kagan et al., 1990; Klein, 1981; Ohman & Mineka, 2001)and are consistent with the suggestion that fear and human thig-motaxis are closely related. Individuals with high measured levelsof fear overestimated the distance from the target to the border ofthe space. Consequently, fearful people have difficulty with iden-tifying the location of the target accurately. On the surface, itappears that the presence of fear and the accompanying thigmo-taxis disrupts either the formation or use of the spatial represen-tation of the current environment.

Although it seems to be plausible that thigmotaxis is a defensivestrategy, it is less likely that thigmotaxis prevents the participant fromlearning the location of the target. Instead, we would argue that fear,related to an encounter with an enclosed spatial environment, triggersa specific exploratory strategy such as thigmotaxis, which plays anessential preparatory role in the first phase of spatial learning. The useof thigmotaxis helps the individual define the borders of an enclosedspace and identify escape routes from that space. Thigmotaxis alsoprovides the individual with the elements of an egocentric frame ofreference. With the elements of that frame of reference in hand, theorganism can begin to construct a cognitive map.

By this view, thigmotaxis impairs spatial learning only when itsuse is prolonged. If prolonged use is related to behavioral inflex-ibility or an inability to switch to an appropriate search strategy,then and only then will spatial learning be disrupted. One mightthen consider behavioral rigidity as being at least partially relatedto a low fear threshold. Once the critical level of fear is reached,gestalt learning will be disrupted by inadequate emotional andperceptual signals.

Thigmotaxis has two functions, each of which relates to differ-ent neuronal structures. During the first phase of learning, theorganism calibrates an enclosed space by using thigmotaxis as anexploratory strategy. Functionally, this defines the borders of thespace and identifies escape routes. In addition, the organism usesthe strategy to gather necessary information for an egocentricframe of reference that provides a context for additional learning.Several authors (e.g., Fanselow, 2000; O’Reilly & Rudy, 2001)have described the dynamics of a transformation from the initialencoding of fragmented, contextual stimulus elements into a uni-fied or gestalt representation and, notably in this context, havedescribed the role of the dorsal hippocampus and the prefrontal

cortex in this transformation. Our results demonstrate that pro-longed thigmotaxis indicates a longer preparation to learn thecontext in a novel environment and that this prolonged activity isassociated with reduced working memory and a reduced capacityto form a general spatial gestalt in normal humans.

Finally, our results are consistent with some neurobehavioralfindings demonstrating that the hippocampus, in both vertebrateand invertebrate animals, is involved in thigmotaxis (Belzung,1992; Besson & Martin, 2004; Kallai et al., in press; Simon et al.,1994). However, the exact neural mechanisms have not been yetfully understood.

Spatial ability is not a single unitary skill; instead, it consists ofsubcomponents. Its main parts are spatial visualization, spatial orien-tation, and the collection and recognition of spatial relations amongproximal or distal cues. Although some researchers (Stumpf & Eliot,1995) suggest that the different spatial functions may be categorizedas a general spatial factor, recent (Quasier-Pohl, Lehmann, & Eid,2004) and earlier (Siegel, 1981) data suggest that spatial ability andthe spatial cognition of environment (cognitive map construction) areindependent. Performance on simple paper-and-pencil tasks does notprovide data leading to a resolution of such topics. For a deeperunderstanding of spatial cognition, the field needs experimental tasksthat require people to use spatial cognition itself. The investigation onspatial orientations through the use of a computer-generated virtualenvironment and digitally recorded and analyzed experiments inlarge-scale real spaces (e.g., Skelton, Bukach, Laurance, Thomas, &Jacobs, 2000; Thomas, Hsu, Laurance, Nadel, & Jacobs, 2001; Wil-son, Foreman, & Stanton, 1997) provides successful empirical ap-proaches for this problem.

The present study analyzed the role of thigmotaxis in spatialorientation in both real and virtual environments. We suggest thatthigmotaxis is a basic element of spatial cognition and emotionallyguided, safety-seeking behavior. On the basis of our results, wesuggest that humans do not only learn the location of objects but thatthey gradually develop an economic and individual structure of thecurrent environment, one that contains further subjective elements.The environment and the person then function as a unit. Acquisitionof a mental map involves spatial knowledge as well as orientations,strategies, cue usage, and abstract and perceptuo–motor processes. Acognitive map is classically defined as a representation of a set ofconnected places systematically related to each other by a group ofspatial transformation rules (O’Keefe & Nadel, 1978). Our resultsdemonstrated how one part of this system, the thigmotaxis strategy,might be able to cooperate with the whole.

Human thigmotaxis, which appears when exploring a novelenclosed space, seems to be related to subjective levels of fear.Functionally, thigmotaxis appears to permit the organism tolearn the borders of the space and to detect escape routes.Thigmotaxis also appears to provide information that can beused as a frame of reference in which spatial mapping mightoccur. This takes place during the first phases of place learning.In the middle phases of place learning, action and knowledgesystems appear to work together to produce a relatively coher-ent map or gestalt of the space. This occurs when the individualuploads landmark-based route information to a knowledge-based map and, conversely, translates map knowledge intomovement of the head and trunk. This translation processrequires broad working memory capacity. If the capacity of

28 KALLAI ET AL.

working memory is overloaded, thigmotaxis may remain unin-hibited and learning may be thereby disrupted.

References

Arrindell, W. A. (1993). The fear of fear concept: Evidence in favor ofmultidimensionality. Behaviour Research and Therapy, 31, 5–18.

Barnett, S. A. (1968). The rat: A study in behaviour. Chicago: Aldine.Belzung, C. (1992). Hippocampal mossy fibers: Implication in novelty

reaction or in anxiety behavior. Behavioural Brain Research, 51, 149–155.

Besson, M., & Martin, J. R. (2004). Centrophobism/thigmotaxis, a newrole for the mushroom bodies in Drosophilia. Journal of Neurobiology,62, 386–396.

Bryant, K. J. (1997). Geographical/spatial orientation abilities within realworld and simulated large scale environments. Multivariate BehavioralResearch, 26, 109–136.

Cheng, K. (1986). A purely geometric module in the rat’s spatial repre-sentation. Cognition, 23, 149–178.

Choleris, E., Thomas, A. W., Kavaliers, M., & Prato, F. S. (2001). Adetailed ethological analysis of the mouse open field test: Effects ofdiazepam, chlordiazepoxide, and an extremely low frequency pulsedmagnetic field. Neuroscience & Biobehavioral Reviews, 57, 253–260.

Creed, R. P., Jr., & Miller, J. R. (1990). Interpreting animal wall-followingbehaviour. Experientia, 46, 758–761.

Devan, B. D., McDonald, R. J., & White, N. M. (1999). Effects of medialand lateral caudate-putamen lesions on place- and cue-guided behav-iours in the water maze: Relation to thigmotaxis. Behavioural BrainResearch, 100, 5–14.

Evans, G. W., Skorpanich, M. A., Ganling, T., Bryant, K. J., & Bresolini,B. (1984). The effects of pathway configuration, landmarks and stress onenvironmental cognition. Journal of Environmental Psychology, 4, 323–335.

Fanselow, M. S. (2000). Contextual fear, gestalt memories, and the hip-pocampus. Behavioural Brain Research, 110, 73–81.

Fraenkel, G. S., & Gunn, D. L. (1961). The orientation of animals: Kineses,taxes and compass orientation. New York: Dover.

Gallistel, C. R. (1990). The organization of learning. New York: MITPress.

Graziano, A., Petrosini, L., & Bartoletti, A. (2003). Automatic recognitionof explorative strategies in the Morris water maze. Journal of Neuro-science Methods, 130, 33–44.

Hartley, T., Maguire, E. A., Spiers, H. J., & Burgess, N. (2003). Thewell-worn route and the path less traveled: Distinct neural bases of routefollowing and wayfinding in humans. Neuron, 37, 877–888.

Jacobs, W. J. (1997). C-G arena [Computer software]. Retrieved fromhttp://w3.arizona.edu/�arg/data.html

Jacobs, W. J., Laurance, H. E., & Thomas, K. G. F. (1997). Place learningin virtual space: I. Acquisition, overshadowing, and transfer. Learningand Motivation, 28, 521–541.

Jacobs, W. J., & Nadel, L. (1999). The first panic attack: A neurobiologicaltheory. Canadian Journal of Experimental Psychology, 53, 92–107.

Jeanson, R., Blanco, S., Fournies, R., Deneubourg, J.-L., Fourcassie, V., &Theraulaz, G. (2003). A model of animal movements in a boundedspace. Journal of Theoretical Biology, 225, 443–451.

Kagan, J., Reznick, J. S., Snidman, N., Johnson, M. O., Gibbons, J.,Gerstein, M., et al. (1990). Origins of panic disorders. In J. C. Ballanger(Ed.), Neurobiology of panic disorder (pp. 71–87). New York: Wiley.

Kallai, J., Kosztolanyi, P., Osvath, A., & Jacobs, W. J. (1999). Attentionfixation training: Training people to form cognitive maps help to controlsymptoms of panic disorder with agoraphobia. Journal of BehaviorTherapy and Experimental Psychiatry, 30, 273–288.

Kallai, J., Makany, T., Csatho, A., Karadi, K., Horvath, D., Kovacs, N., etal. (in press). Thigmotaxis navigation strategy and hippocampus volu-

metry: A study with Morris type mazes and the neurobehavioral corre-lates of spatial strategies. Behavioural Brain Research.

Kallai, J., Makany, T., Karadi, K., & Jacobs, W. J. (2005). Spatial orien-tation strategies in Morris-type virtual water task for humans. Behav-ioural Brain Research, 159, 187–196.

Klein, D. F. (1981). Anxiety re-conceptualized. In D. F. Klein & J. G.Rabkin (Eds.), Anxiety: New research and changing concepts (pp.8–41). New York: Ravel Press.

Kozlowski, L. T., & Bryant, K. J. (1997). Sense of direction, spatialorientation, and cognitive maps. Journal of Experimental Psychology:Human Perception and Performance, 3, 590–598.

Kun, M., & Szegedi, M. (1971). Az intelligencia merese [The assessmentof intelligence]. Budapest, Hungary: Akademiai Kiado.

Lawton, C. A. (1994). Gender differences in way-finding strategies. SexRoles, 30, 765–779.

Lipp, H. P., & Wolfer, D. P. (1998). Genetically modified mice andcognition. Current Opinion in Neurobiology, 8, 272–280.

McDonald, R. J., & White, N. M. (1994). Parallel information processingin the water maze: Evidence for independent memory systems involvingdorsal striatum and hippocampus. Behavioral and Neural Biology, 61,260–270.

Morris, R. G. M. (1984). Developments of a water-maze procedure fromstudying spatial learning in the rat. Journal of Neuroscience Methods,11, 47–60.

Noldus, L. P. J. J., Spink, A. J., & Tegelenbosch, R. A. J. (2001). Aversatile video tracking system for automation of behavioural experi-ments. Behavior Research Methods, Instruments, & Computers, 33,398–414.

Ohl, F., Sillaber, I., Binder, E., Keck, M. E., & Holsboer, F. (2001).Differential analysis of basal behaviour and diazepam-induced alter-ations in C57Bl/6 and BALB/c mice using the modified hole board.Journal of Psychiatry Research, 35, 147–154.

Ohman, A., & Mineka, S. (2001). Fears, phobias, and preparedness:Toward an evolved module of fear and fear learning. PsychologicalReview, 108, 483–522.

O’Keefe, J., & Nadel, L. (1978). The hippocampus as a cognitive map.Oxford, United Kingdom: Clarendon Press.

O’Reilly, R. C., & Rudy, J. W. (2001). Conjunctive representations inlearning and memory: Principles of cortical and hippocampal function.Psychological Review, 108, 810–817.

Pellow, S., & File, S. E. (1986). Anxiolytic and non anxiogenic drug effectson exploratory activity in an elevated plus-maze: A novel test of anxietyin the rat. Pharmacology Biochemistry and Behavior, 24, 525–529.

Quasier-Pohl, C., Lehmann, W., & Eid, M. (2004). The relationship be-tween spatial abilities and representations of large-scale space in chil-dren: A structural equation modeling analysis. Personality and Individ-ual Differences, 36, 95–107.

Reid, A. K., & Staddon, J. E. R. (1998). A dynamic route finder for thecognitive map. Psychological Review, 105, 385–401.

Reid, R. A., & Reid, A. K. (2005). Route finding by rats in an open arena.Behavioural Processes, 68, 51–67.

Schmitz, S. (1997). Gender related strategies in environmental develop-ment: Effect of anxiety on way finding in the representation of athree-dimensional maze. Journal of Environmental Psychology, 17,215–228.

Siegel, A. W. (1981). The externalization of cognitive maps by childrenand adults: In search of ways to ask better questions. In L. S. Liben,A. H. Patterson., & A. N. Newcombe (Eds.), Spatial representations andbehaviour across the life span (pp. 167–194). New York: AcademicPress.

Simon, P., Dupuis, R., & Constentin, J. (1994). Thigmotaxis as an index ofanxiety in mice: Influence of dopaminergic transmission. BehaviouralBrain Research, 31, 59–64.

29THIGMOTAXIS STRATEGY IN HUMANS

Sipos, K. (1978). Spielberger State–Trait Anxiety Inventory magyar val-tozata [Spielberger State–Trait Anxiety Inventory—Hungarian version],Budapest, Hungary: MTA Press.

Skelton, R. W., Bukach, C. M., Laurance, H. E., Thomas, K. G. F., &Jacobs, W. J. (2000). Humans with traumatic brain injuries show place-learning deficits in computer-generated virtual space. Journal of Clinicaland Experimental Neuropsychology, 22, 157–175.

Spielberger, C. D., Gorsuch, R. L., & Lushene, R. E. (1970). State–TraitAnxiety Inventory Test manual for Form 1. Palo Alto, CA: ConsultingPsychologist Press.

Stumpf, H., & Eliot, J. (1995). Gender-related differences in spatialability and the k factor of general spatial ability in a population ofacademically talented students. Personality and Individual Differ-ences, 19, 33– 45.

Thomas, K. G. F., Hsu, M., Laurance, H. E., Nadel, L., & Jacobs, W. J.(2001). Place learning in virtual space: III. Investigation of spatialnavigation training procedures and their application to fMRI and clinicalneuropsychology. Behavior Research Methods, Instruments, & Comput-ers, 33, 21–37.

Treit, D., & Fundytus, M. (1988). Thigmotaxis as a test for anxiolyticactivity in rats. Pharmacology Biochemistry and Behavior, 31, 959 –962.

Waller, D. (2000). Individual differences in spatial learning from computersimulated environments. Journal of Experimental Psychology: Applied,6, 307–321.

Wang, R. F., & Spelke, E. (2004). Comparative approaches to humannavigation. In K. J. Jeffery (Ed.), The neurobiology of spatial behavior(pp. 119–143). New York: Oxford University Press.

Wechsler, D. (1997). Wechsler Adult Intelligence Scale—3rd Edition(WAIS–III). San Antonio, TX: Harcourt Assessment.

Whishaw, I. Q., Hines, D. J., & Wallace, D. G. (2001). Dead reckoning(path integration) requires the hippocampal formation: Evidencefrom spontaneous exploration and spatial learning task in light(allothetic) and dark (idiothetic). Behavioral Brain Research, 127,49 – 69.

Wilson, P. N., Foreman, N., & Stanton, D. (1997). Virtual reality andrehabilitation. Disability and Rehabilitation, 19, 213–220.

Appendix

Arena Instruction

In this experiment, your task is to find a target on the floor of acomputer-generated room. You will be transported into two dif-ferent computer-generated rooms, a practice room and an experi-mental room. Both rooms contain circular arenas inside largesquare rooms.

You will start in a practice room. It has brightly colored walls,a white ceiling, a red arena wall, and a gray floor. In this room, allyou need to do is practice moving around using a joystick.

Moving and Looking

To go forward, push the joystick forward.To go backward, pull the joystick backward.To turn to the right, push the joystick right.To turn to the left, push the joystick left.Remember: pushing the joystick left or right will turn you in the

corresponding direction, but will not move you sideways.When you have mastered moving and looking, press the space

bar on the keyboard and you will be transported to the experimen-tal room.

In the experimental room, your task is to search for, find, andstand on a large blue target. The first two trials of the experimenthave visible targets. With a visible target, simply move to it as

quickly as you can. You will know you are on the target when youhear a drum. Once you’re on the target, you won’t be able to getoff, so just hit the space bar to teleport to the waiting room.

The next 8 trials of the experiment will have an invisible target.With the invisible target, you should search the room to find it. Theinvisible target is always in the same place, so you should take aGOOD LOOK around the room when you find it. When you havetaken a good look around, press the space bar to go back to thewaiting room.

There will be a trial with a visible target at the end of theexperiment. Again, for the visible target, move to it as quickly asyou can.

On each trial, you will have limited time to find the target. If yougo over this time you will be automatically transported back to thepractice room.

Remember: the visible targets will be in different places, but theinvisible target will always in the same place, so take a good lookaround when you first find the invisible target.

Received March 23, 2006Revision received September 6, 2006

Accepted September 27, 2006 �

30 KALLAI ET AL.